In addition to the substantial challenges of planning, coordinating, and implementation, the international response to AIDS poses significant fiscal challenges. It amounts to a substantial commitment of financial resources by governments or donors and frequently takes place in the context of weak health sector capacities. In this article, we first summarize key aspects of the financing of AIDS programmes. Second, we discuss the fiscal dimension of the impact of and the response to AIDS, which arise as a government's resource envelope tightens while the demands on government grow. Third, we review to what extent increased external financing has translated into increased access to treatment.
Financing the response to AIDS: an overview
On a global scale, AIDS-related spending has increased very substantially over past years. Joint United Nations Programme on AIDS (UNAIDS) estimates that spending in low and middle-income countries increased from approximately US$300 million in 1996 to approximately US$ 8.3 billion in 2005, and projects a further increase to US$10 billion for 2007 . Most of the increase is accounted for by external grants from bilateral or multilateral sources, which amounted to approximately 85% of public spending on AIDS in 2005, whereas 15% of the costs were financed from domestic government budgets. This increase in AIDS-related external financing over past years is also significant when seen in the context of total aid flows. In 2005, total aid flows from Organization for Economic Cooperation and Development (OECD) countries (excluding debt relief for Nigeria and Iraq) amounted to approximately US$85 billion, which means that AIDS-related programmes correspond to approximately 8% of all OECD development assistance .
These aggregate numbers of AIDS-related spending mask very substantial differences across countries and the extent to which the national response in many countries is dependent on external aid. Most notably, a few middle-income countries account for the bulk of domestically financed spending. According to UNAIDS, domestic spending on AIDS amounted to US$115 million in Argentina, US$165 million in Botswana, US$386 million in Brazil, US$99 million in China, and US$447 million in South Africa in 2005, and it had attained US$196 million in Mexico in 2002 . These five countries account for approximately two-thirds of all domestically financed spending on AIDS recorded by UNAIDS, whereas grants account for well over 90% of AIDS-related spending in low-income countries.
Table 1[1–6] summarizes the available data on the financing of AIDS-related expenditures in some of the worst affected countries. The table does not show funding by private institutions and thus may understate external financing on AIDS programmes in some countries. We find that AIDS-related expenditures account for between 0.3 and 4.2% of gross domestic product (GDP). Of the 13 countries covered, AIDS-related spending exceeds one-quarter of public health expenditure in nine countries, and one-half of health expenditures in two. (Although such comparisons are useful to illustrate the scale of the response to AIDS in a national context, it is important to note that some of the AIDS-related expenditures financed by grants should not be classified as health expenditures, but as educational or social spending.)
Even more than it is the case for overall health expenditure, the expansion of AIDS programmes is driven by external financing, which accounts for over 90% of AIDS-related spending in five of the 13 countries, and for over 80% in nine countries. (The other four countries include three middle-income countries and Zimbabwe.) Table 1 also illustrates the key role played by the Global Fund to Fight AIDS, Tuberculosis and Malaria and the US President's Emergency Plan for AIDS Relief in Financing AIDS programmes: each one accounts for more than half of total disbursements in three of the countries in Table 1.
Fiscal impact of AIDS
In terms of the fiscal dimension of the national response to AIDS, it is useful to distinguish two perspectives. One focuses on the costs of AIDS-related programmes, most directly as the funding of a national programme or specific projects (e.g. in a grant proposal). The other focuses on the implications of AIDS (including the response) for the government's overall resource envelope. In addition to the costs of AIDS programmes, and, for many countries, grants that at least partly offset the fiscal costs of AIDS-related programmes, the most important aspects of this fiscal policy perspective on AIDS are any impacts of AIDS on the capacities of public services, costs associated with increased morbidity and mortality among public servants, certain social expenditures, and the implications of AIDS for government revenues and debt sustainability.
Part of the fiscal impact of AIDS arises through increased mortality among government employees, which, most obviously, erodes government capacities directly with lost skills and greater attrition. Higher morbidity and mortality also carry certain financial costs, including medical benefits, death and funeral-related benefits, and training and recruitment costs. Available estimates of these costs suggest that in a setting with an HIV prevalence of approximately 20%, funeral grants could increase personnel costs by approximately 0.5%, and the (financial or efficiency) costs of increased absenteeism could amount to approximately 2–3% of the working time . Medical expenses can add substantially to these costs (while reducing other costs); however, when treatment is delivered through grant-financed health programmes, the impact on the fiscal balance could be much smaller than the total costs.
The changed mortality pattern and demographic profile also affect the benefit incidence of pension and social insurance schemes . Here, the fiscal impact is ambiguous, as the increase in death-related benefits and pensions to surviving dependents is partly offset by a decline in the number of people reaching retirement age as well as lower survival rates of surviving partners. The impact of AIDS on the balance of pension systems (and, if the pension fund is operated or owned by the government, the fiscal balance) also depends on the extent to which the risk of premature retirement or death is borne by the contributor (e.g. if the scheme simply pays out accumulated contributions) or by the pension or social insurance fund (e.g. lump sum funeral grants).
Adequate data on the impact of AIDS on government revenues are not available. Many components of the tax base such as aggregate payroll, company profits, or imports are, however, related to the level of GDP. Whereas the evidence on the impact of HIV/AIDS on the growth or level of GDP per capita is mixed, a slowdown in population growth would nevertheless result in a decline in overall GDP growth, suggesting that government revenues also grow at a reduced rate [7,9].
As AIDS results in a slowdown in GDP growth, it also has implications for the sustainability of public debt. The most common indicator of public debt is the ratio of debt (D) to GDP. Suppose that AIDS results in a slowdown in GDP growth from 3 to 2%, and that the government aims to stabilize debt at 40% of GDP, so that D/GDP is constant. This scenario would require an adjustment in the fiscal balance of 0.4% of GDP. The government's fiscal strategy may be more complex, and could initially include an increase in domestically financed spending. Nevertheless, in the context of reduced growth, any path for the fiscal deficit results in a faster accumulation of debt, and the government would eventually have to find ways to adjust the fiscal deficit.
In the same way that the fiscal impact of AIDS goes beyond the costs of the national response to AIDS, the national response also has wider fiscal repercussions, as a successful response to AIDS partly reverses some of the economic and fiscal effects referred to earlier. Masha , for example, estimates that fiscal savings in other sectors associated with Botswana's national strategic framework on AIDS amount to approximately 1% of GDP (14% of the costs of the national strategic framework).
To summarize, in addition to the costs associated with the response to the epidemic, AIDS narrows the government's fiscal space by increasing personnel costs and other fiscal outlays associated with increased morbidity and mortality among the population, as well as modestly reducing government revenues. At the same time, the government's capacities are eroded through increased attrition among public servants (which can be mitigated, but not fully offset, through treatment). Looking ahead, the reduced growth outlook may also require a more cautious fiscal policy stance, in light of the dynamics of debt accumulation.
Macroeconomic and capacity constraints to rapid scaling up
The international response to the epidemic involves a rapid scaling up of AIDS-related expenditures. In this section, we will first discuss whether the speed or the scale of the response could be subject to capacity constraints that would cause adverse macroeconomic repercussions, by placing it in the context of overall aid flows. Second, we will discuss the effectiveness of efforts to increase access to treatment, based on the data on access to treatment from World Health Organization (WHO)/UNAIDS/United Nations Children's Fund (UNICEF) .
The issue of aid effectiveness and the macroeconomic effects of large aid flows has gained new prominence in the context of recent commitments to attaining the millennium development goals. Such potential macroeconomic effects include the following: (1) an aid-financed increase in the domestic demand for goods and services could result in an increase in domestic prices, i.e. higher inflation. (2) To the extent that aid translates into domestic spending (rather than imports), the supply of foreign currency increases relative to the demand, resulting in an appreciation of the domestic currency, with adverse effects on the country's competitiveness (alternatively, the central bank may buy the foreign exchange, resulting in an increase in the money supply and inflation) .
The question as to what extent these macroeconomic considerations apply to the financing of AIDS programmes is an empirical one. Table 2 provides some insights regarding the scale of AIDS-related aid flows from a macroeconomic perspective. For five of the countries shown, AIDS-related aid flows exceed 1% of GDP; in two cases they exceed 2%. Whereas these amounts are significant from a fiscal perspective, they are dwarfed by the overall volume of external aid, which amounts to approximately 20% of GDP in some of the countries shown. Whereas AIDS-related aid flows account for a substantial proportion of incoming aid in some countries, most notably in southern Africa, these countries tend to receive relatively low amounts of aid overall. Therefore, given the scale of externally financed AIDS programmes in the broader context of external aid, it appears very unlikely that AIDS-related aid flows would create macroeconomic issues such as those hypothesized above.
Although AIDS-related aid flows appear unlikely to cause macroeconomic imbalances by themselves, one characteristic of AIDS spending is that most is concentrated in the health sector, which raises the possibility that limited (pre-)existing capacities may constrain the pace or the effectiveness of an expansion in AIDS-related health services. In this regard, Fig. 1 plots rates of access to antiretroviral treatment (ART) against HIV prevalence rates (as a measure of the scale of the health challenge). Three preliminary lessons emerge: (1) high HIV prevalence rates are not an insurmountable obstacle in terms of access to treatment; many countries with high prevalence rates also feature high rate of access. (2) With the notable exception of Malawi and Zambia, however, the countries facing a high HIV burden that are successful in providing treatment are middle rather than low-income countries (Botswana, Namibia, South Africa and Swaziland). (3) A large number of low-prevalence countries have low rates of access to treatment, which may point to political rather than capacity constraints as obstacles to high rates of treatment access.
There may be several potential explanations for the observed differences in access to treatment across countries. We will focus on the following: (1) The scale of the epidemic (contrary to the first impression provided by Fig. 1) does have a bearing on the success of expanding access to treatment. (2) Wealthier countries, and residents of these countries, are in a better position to afford increased access to treatment. (3c) The scope for increasing access to treatment may be constrained by (pre-)existing health sector capacities. (4) Domestic spending on AIDS and the extent to which countries benefit from external assistance may differ.
To address these possibilities, we have constructed a dataset including data on coverage rates of ART, HIV prevalence and domestic spending on AIDS (from UNAIDS) , ART coverage rates and birth attendance rates (the latter as a measure of the capacities of the health sector) from WHO, UNAIDS, and UNICEF , the numbers of doctors and nurses , population size , GDP and GDP per capita , and external financing . For our analysis, we transformed the data on external aid and domestic spending into US dollars per person living with AIDS in each country. This yielded a dataset of 87 countries for which all of these variables were available.
As our dependent variable is censored (it can take values between 0 and 100 only), we apply censored normal TOBIT maximum likelihood estimators throughout. In addition to the R2 of the regression, we show the t-statistics pertaining to the respective coefficients (in brackets); 1, 2, or 3 stars indicate coefficients significant at the 10, 5, and 1% confidence level. Before presenting our broader analysis, we return to our interpretation of the data shown in Figure 1. We find that the HIV prevalence rate (HIV) alone explains almost none of the variations in coverage rates of ART (TREAT) in equation 1, which returns an R2 of only 0.001. Adding GDP per capita (equation 2) raises the R2 to 0.36, which means that differences in GDP per capita are also associated with much of the variations among the low-prevalence countries in Figure 1.
To gain a better understanding of the different factors that may have a bearing on treatment coverage rates and the validity of the associated hypotheses, we have included all of the variables mentioned above in our regressions. Eliminating successively insignificant variables (the numbers of doctors, nurses, and domestic expenditure on AIDS), we arrive at equation 3:
which explains access to treatment in terms of HIV prevalence, GDP per capita, external aid per capita (AIDPC), and birth attendance rate (BIRTHATT). We find that an increase in HIV prevalence translates into a treatment coverage rate that is lower by 1.1 percentage points; a difference in GDP per capita of US$1000 is associated with a difference in treatment access of 7 percentage points; US$1 in external aid per capita is associated with higher treatment acccess of 2.9 percentage points; and a birth attendance rate that is one percentage point higher translates into a treatment coverage rate that is 0.3 percentage points higher.
Some notes are in order about the interpretation of these results. The first lesson is that in terms of expanding access to ART, the economic context matters. Treatment coverage rates have reached higher levels in countries where income per capita is higher and which feature a higher coverage rate of basic health services. Second, we find that HIV prevalence now enters negatively and significantly, sensibly suggesting that– controlling for other factors– it is easier to attain high treatment coverage rates when HIV prevalence is low. In particular, the fact that HIV prevalence was insignificant in the preliminary regressions but enters significantly in Eq. (3) can be attributed to the inclusion of external aid in this and other regressions. One plausible interpretation is that the international response to AIDS has been effective in levelling the field in terms of access to treatment between high and low-prevalence countries. Third, not only inputs, but also the effectiveness of health services is an important determinant of access to treatment. Health inputs alone as indicators of health sector capacities explain little of the variation in access to treatment, whereas we see higher rates of access to treatment when health inputs translate into positive health outcomes (here measured by birth attendance rates).
Some fairly substantial limitations to our analysis remain. First, we cannot distinguish between different types or purposes of AIDS-related aid. Our data include low-prevalence countries where a larger share of aid is spent on prevention efforts, and high-prevalence countries where a larger share of funds is spent on treatment. As a consequence, we also cannot meaningfully use aid per person living with AIDS as a potential determinant of access to treatment, as this is highest in low-prevalence countries where it primarily reflects prevention efforts. Second, our dataset did not allow us to convincingly assess whether there are interaction effects between aid and the quality of health services. For example, in terms of significance or goodness-of-fit, there was little basis for choosing between AIDPC and AIDPC*BIRTHATT in many regressions. Third, much of the variation in access to treatment among low-prevalence countries does not reflect feasibility constraints, and it would be desirable to obtain better measures for constraints for these countries, for example in the area of political economy.
In conclusion, we find little basis for concerns about macroeconomic constraints to scaling up the international response to HIV/AIDS, in light of the moderate scale of AIDS-related aid flows relative to overall aid. Regarding sectoral contraints, the picture is more differentiated. Many countries with high prevalence rates have also achieved high rates of access to treatment, although most of these countries are middle-income countries. Our econometric analysis credits external aid as a key factor that has enabled higher-prevalence countries to cope with the additional demands for health services. At the same time, GDP per capita and health sector capacities are important determinants of access to treatment.
Disclaimer: The views presented here are the author's and do not necessarily represent those of the IMF or its Executive Board.
Conflicts of interest: None.